The Rb gene was isolated almost 20 years ago, but fundamental questions regarding its role in retinal development and retinoblastoma remain. What is the normal function of RB protein in retinogenesis? What is the cell-of-origin of retinoblastoma? Why do retinoblastoma tumors have recurrent genetic lesions other than Rb inactivation? Why is retinoblastoma not induced by defects in cell cycle regulators other than Rb? Why is the retina so sensitive to Rb loss? Recently developed conditional Rb knockout models provide new insight into some of these issues. The data suggest that RB protein may not control the rate of progenitor division, but is critical for cell cycle exit when dividing retinal progenitors differentiate into postmitotic transition cells. This finding focuses attention on the ectopically dividing transition cell, rather than the progenitor, as the cell-of-origin. Cell-specific analyses in the RB-deficient retina reveal that ectopically dividing photoreceptors, bipolar and ganglion cells die, but amacrine, horizontal and Muller cells survive and stop dividing when they terminally differentiate. Rare amacrine transition cells escape cell cycle exit and generate tumors. These data suggest that post-Rb mutations are required to overcome growth arrest associated with terminal differentiation, rather than apoptosis as previously suggested. To explain why perturbing cell cycle regulators other than RB does not initiate retinoblastoma, we speculate that mutations in other components of the RB pathway perturb cell cycle arrest, but only RB loss triggers genome instability in retinal transition cells, which may be critical to facilitate post-Rb mutations necessary for transformation. Cell-specific differences in the effect of Rb loss on genome stability may contribute to the tremendous sensitivity of retinal transition cells to tumorigenesis. The new mouse models of retinoblastoma will be invaluable for testing these possibilities.